Image or object recognition is key for a autonomous robot like Aido. Aido needs to identify items when navigating and decide real time how to avoid them when it’s moving. We scouted around for software to help with this, and chose Uncanny vision’s offline image recognition library to power Aido’s image recognition.
Uncanny Vision is a technology firm based out of Japan and India that has over 25 years of experience in the Computer Vision area. Some key advantages are:
- Uncanny Vision’s image recognition library recognizes over 3000 objects
- The library runs offline with a low memory footprint. Aido is able to recognize objects up to 10x faster than conventional methods using SIMD instructions. Offline recognition also cuts out network latency and makes the image recognition more stable.
- Another advantage of the low latency is that it allows Aido to do a frame-by-frame comparison of visual input. Using this, Aido is able to figure out if there are moving objects in the frame. If these are in the path it has to traverse, Aido stops to avoid a collision.